Crowding-in and crowding-out effects of public investments in the Portuguese economy

Abstract

This study analyses the effects of public and private investment on Portuguese GDP in the period 1960–2013. After a review of the literature based on works developed in the context of VAR analyses, an alternative econometric strategy is proposed. We opt for ADL models using the methodology of Krolzig-Hendry-Doornik. We estimate direct effects of public and private investments on themselves and also a system of simultaneous equations and calculate the multipliers of the exogenous variables, represented by the current external transfers, the short-run nominal interest rate and public debt ratio. Additionally, we tested a model with the first three equations of the system, using beyond those variables the real exchange rate as an exogenous variable. The results point to the existence of a complementarity between private investment and public investment rather than any idea of substitutability. We find that public debt has important negative effects on public and private investments and consequently on output. Public investment has positive effects on output and on private investment. The appreciation of the real exchange rate has an important and long-lasting negative effect upon output, confirming the presence of a mechanism associated with a Dutch-disease phenomenon in the Portuguese economy.

Keywords:

• crowding-in
• crowding-out
• Portuguese economy
• private investment
• public investment
• ADL and 3SLS

JEL Classifications:

• C32
• E22
• E62

Acknowledgments

The authors thanks two anonymous referees for their very helpful comments and suggestions.

Notes

1. In the first column we have the explanatory variables; in the second the coefficient values; in the third the standard-errors associated with the estimated coefficients; in the fourth the T values; in the fifth the probability values associated with the T; and finally in the last column some other statistical information, the standard error of the estimation, the Hansen J-specification test and its order, the over-identification test of instruments as an LM test and its order, the Breusch-Godfrey LM test for auto-correlation of order 1 to 2, and the ADF tests with the t and the Z statistics usually the last tests. In this table two other tests are presented: for the null sum of two regression coefficients. If not indicated the tests are always Chi-squared tests.